Chapter 3 Low frequency dielectric measurements

3.1 Introduction
After about sixty years of lingering development 3, 211], low-frequency (LF) dielectric measurements are nally emerging as an important materials characterization tool. In the last few years, they have become indispensable in such diverse elds as: mining industry | for monitoring the quality of coal 212]; pharmacological industry | for estimating adsorption of drugs onto nanoparticles for novel drug delivery systems 213]; polymer industry | for monitoring the kinetics of polymerization in polymerizing systems 214, 215]; oil industry | for estimating the water content in crude oil 216]; biology | for online monitoring of cell concentrations during yeast cultivation and other fermentation processes 217, 218]; 118

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

119

medicine | for non-invasive measurements of tissues and bones, possibly leading to the development of diagnostic techniques for early detection of such maladies as cancer and osteoporosis 219, 220]; material science | for optimization of the cure cycle of composites 3, 221]; and food science | for prediction of the level of cholesterol in certain food ingredients 222]. The LF dielectric measurements are unequaled in their ability to dynamically monitor the many chemical and physical processes important in investigations of new materials | such as cure, polymerization, phase transitions, and di usion. In this Chapter, we shall review the LF dielectric measurement techniques that we used to investigate the dielectric properties of fullerenes. Section 3.2 describes the conventional parallel-plate technique for measuring the dielectric function "(!). Sec. 3.3 deals with theoretical and practical considerations of microdielectrometry. Finally, Sec.3.4 discusses special techniques for high-impedance transport measurements. In addition, Appendix A contains the LabView code for the software written to control the measurement systems and instruments described in this section.

3.2 Parallel-plate measurements

3.2.1 Overview
The conventional way for making low frequency (below 1 MHz) measurements of the dielectric properties of solids is to place a sample between closely spaced parallel conducting plates (Fig. 3-1), and to monitor the AC equivalent capacitance Cm(!) and dissipation factor (also known as loss tangent ) D(!) of the resulting capacitor. Normally, one designs the plate spacing to be much less than the plate size, as this serves to minimize the e ects of fringing eld.

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

120

Figure 3-1: Parallel plate electrodes. The material under test in the parallel-plate con guration (Fig. 3-1) can be modeled as a frequency-dependent capacitance C(!) in parallel with a frequencyindependent resistor R0 Fig. 3-2(a)]. The DC resistance R0 takes into account processes such as tunneling, thermally activated hopping, and ionic conduction. The capacitance C(!) is proportional to the complex dielectric function "(!) = "0(!) j"00(!) of the sample material under test:

C(!) = "(!)"0A=d;

(3.1)

where A is the cross sectional area of the capacitor, d is the separation between the plates, and "0 = 8:85 10 12 F/m is the absolute permittivity of free space. The capacitance Cm (!) measured by a typical capacitance bridge (such as an HP 4274A by Hewlett Packard, Inc. that was used in this study) is equal to the real part of C(!): Cm(!) = "0(!)"0A=d; (3.2) and so it can be used to extract the frequency-dependent real part of the dielectric function "0(!). The other measured quantity, the dissipation factor (the loss tangent )

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

given by

121 (3.3)

D(!) = "00(!)="0(!) + !" "0(d )AR ; !

0 0

can then be used to extract the imaginary part of the dielectric function "00(!).

3.2.2 Special considerations

While the theory underlying the parallel-plate technique described above is quite simple, an actual measurement is complicated by a variety of additional constraints arising from the properties of the material being investigated, the nature of the interface, and the limitations of the measurement system. These three classes of constraints, as applied to the measurement of C60 solid lms, are brie y discussed below.

Interfacial e ects
In most actual measurements, the e ect of interfacial polarization giving rise to the interfacial capacitance Ci needs to be taken into account. For a typical solid, the physics behind Ci includes: 1. The purely interfacial electronic e ects in uenced by the presence of electrons trapped with a nite lifetime in localized states near the interface 223], and 2. The \blocking layer" space-charge polarization e ects due, e.g., to the presence of mobile ions in the material 3, 193, 224] (see Fig. 3-3). The measured capacitance Cm is then equal to the series combination of the interface capacitance Ci and the sample (bulk) capacitance C 225], as shown in Fig. 3-2(b). One thus expects the total reciprocal areal capacitance to have a linear dependence on thickness d of the sample with a nonzero intercept at d = 0, i.e.,

A =A+ d : Cm Ci "0"0

(3.4)

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

122

+E +() 4o +() 4o

(a)

(b)

Figure 3-2: Equivalent circuits for the material under test in Fig. 3-1 with (b) and without (a) taking the interfacial capacitance into account.

Figure 3-3: Idealized view of the polarization due to ions. Ion conduction in the presence of an applied electric eld leads to electrode (interfacial) polarization and formation of a \blocking layer" (after Day et al. 193]).

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

123

Note that as the thickness d of the lm gets smaller, the interfacial capacitance Ci may come to dominate the measured capacitance Cm. Hebard et al. have shown 225] that, in Al-Al2O3-Al trilayer structures, this happens when the oxide thickness falls below 50 A. We, therefore, expect the bulk capacitance due to C60 to dominate the capacitive part of the impedance in the Al-C60-Al trilayer structures with dC60 1000 A that were used in this study (see Sec. 4.6 and Fig. 4-13).

Linearity and time invariance

In addition to the interfacial e ects, one needs to keep in mind that, in order to be able to reliably interpret a dielectric measurement, it needs to be performed in the linear regime. Whether or not a given sample is linear depends on the applied voltage. Most dielectrics become nonlinear at electric elds on the order of 105 V/cm, and experience catastrophic breakdown at electric elds on the order of 106 V/cm. Therefore, when dealing with lms of thicknesses on the order of 1000 A (a typical lm thickness in our studies), the applied voltage must be kept at 0.1 V or lower to keep a safe distance of two orders of magnitude or more away from the typical breakdown voltage. The other important characteristic of a measured dielectric system is its time invariance. While usually not a problem at frequencies of 1 Hz and higher, this condition is very di cult to ful ll in very low frequency (VLF) measurements, when a single frequency scan may take hours, or even days. Temperature uctuation, humidity, di usion of ambient gases, and sample changes caused by other factors may complicate the interpretation of the results. Therefore, care should be taken to ensure that the sample changes only insigni cantly during the interval required to make a single sample measurement.

Cable capacitance
Additional problems are introduced by cabling and shielding issues. In Fig. 3-4, the

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

124

Figure 3-4: Schematic diagram of a measurement system illustrating the e ect of cable capacitance and shielding. parallel-plate sample is connected to a meter by a coaxial cable ( 100 pF/m). A 2-meter cable introduces a capacitance of 200 pF in parallel with the sample capacitance, which must be subtracted from the measured capacitance Cm to determine the sample capacitance. Clearly, when the cable capacitance is comparable to the sample capacitance, the sensitivity and accuracy are reduced. It is, therefore, advisable to use the twisted-pair cables whenever possible, as they usually possess smaller capacitance per unit length than the coaxial cables. Additionally, the cable and electrodes must be shielded, to prevent the pickup at 60 Hz and its harmonics which degrade measurements at or near those frequencies.

3.2.3 Measurement equipment

Whether using parallel plate or other types of electrodes, an AC measurement involves determining the admittance between the electrodes under sinusoidal steady-state conduction. Any AC measurement system thus includes a generator to produce the sinusoidal excitation fed to the sample, and an analyzer to compare the magnitude and phase of the output with the excitation input. The instruments and con gurations used in this study are described below.

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

125

Figure 3-5: A schematic of the working principle of the SI 1254 Frequency Response Analyzer.

HP 4274A LCZ meter

Hewlett Packard's HP 4274A LCZ Meter is a general-purpose impedance-measuring instrument designed to measure circuit components such as capacitors and inductors in the frequency range from 100 Hz to 100 kHz. The magnitude of the sinusoidal input excitation can be varied from 0.01 V to 10 V, and was usually xed at 0.1 V to ensure that the measurement stays in the linear regime, as discussed above. The instrument is equipped with the standard IEEE 488 computer bus, which facilitates interfacing and control from a lab computer. Appendix A.1 describes the LabView software interfacing the instrument with an IBM-compatible PC.

SI 1254 Frequency Response Analyzer

The Digital Frequency Response Analyzer (FRA) SI 1254 by Solartron Instruments (a division of Schlumberger Ltd.) is an example of the digital correlator system. The principle of transfer function measurement is shown schematically in Fig. 3-5. The sample response S (t) to the perturbing signal x(t) = X0 sin !t is correlated with two

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

126

synchronous reference signals, one in phase with x(t) and the other 90o out of phase, in order to calculate the real and imaginary parts of the sample's transfer function S(!): 1ZT RefS(!)g T S (t) sin !t dt (3.5) (3.6) ImfS(!)g T S (t) cos !t dt 0 where T , the integration time, is an integer number of periods of the perturbing signal. The FRA thus yields directly the real and imaginary parts of the transfer function of the sample, which, in SI 1254, can alternatively be displayed in the phase{magnitude notation. Since the input impedance of the SI 1254 FRA is relatively low (100 M ), the output signal S (t) needs to be channeled through a high-bandwidth ampli er, such as the SR 570 current ampli er described in Sec. 3.4.1 Because the FRA is a digital instrument, it can be used to perform measurements at arbitrarily low frequencies. In this study, we have measured the dielectric properties of C60 lms at frequencies as low as 1 mHz. Appendix A.2 presents the LabView software interfacing the SI 1254 FRA with a personal computer. 1ZT
0

Other equipment
All of the functions of the FRA can alternatively be implemented directly in a modern personal computer equipped with appropriate Input/Output hardware using digital signal processing techniques. Work has been started on a digital computer-based data acquisition system (DAQ) based on the National Instruments' AT-MIO-16E-3 multifunction I/O board. The board features two analog output channels for generation of the excitation signal, and 16 analog input channels for output detection. The 500 kSamples/s, 12-bit (1 in 4,096) resolution, and 100 G input impedance proThe SI 1286 Electrochemical Interface by Solartron Instruments, with input impedance exceeding 10 G , can also be used for signal conditioning.
1

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

127

vide the operational parameters better than those of the SI 1254 FRA. A LabView implementation of the frequency analysis system is currently in progress.

3.3 Microdielectrometry
Clearly, the calibration of the parallel-plate measurements described in Sec. 3.2 will be a ected if the plate spacing d changes in the course of the measurements, as it will, for instance, at the C60 phase transition point T01 260 K (Sec. 1.3.2 and Fig. 1-11), or during intercalation of oxygen and other species into the interstitial spaces of C60 solid (Sec. 4.7), or during photopolymerization of a C60 lm (Sec. 1.3.2). To avoid this problem, we have used the microdielectrometry technique 226{228] to measure dielectric properties of C60 at low frequencies. Instead of using a parallel plate geometry, both electrodes used in the microdielectric measurement are placed on the same surface of an integrated circuit (see Fig. 3-6), and the medium to be studied (C60 lm) is placed over the electrodes by thermal sublimation in vacuum (Sec. 4.5). The comb electrodes in Fig. 3-6, in contrast to parallel plates, provide a xed calibration for both "0 and "00, because electrode size and spacing remain constant even when the sample itself is changing. However, the interdigitated electrode geometry is less e cient than parallel plates in terms of coupling the electric eld through the sample medium. To increase the signal to an easily measurable level, some microdielectrometer sensors provide ampli cation in the form of an integrated eld-e ect transistor, whose gate electrode is one of the two interdigitated electrodes (called the oating gate; see Fig. 3-6). To compensate for the transistor ampli cation factor, which would require a complex calibration procedure at each operating temperature, a second identical eld-e ect transistor is simultaneously fabricated on the same integrated circuit. The second eld-e ect transistor serves as a reference in a di erential feedback circuit, the details of which have been published elsewhere 229]. The net e ect of the two-transistor plus feedback circuit combination is that when

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

128

Figure 3-6: (a) Schematic view of active portion of microdielectrometer sensor chip. CFT refers to \ oating-gate charge- ow transistor," of which the microdielectrometer is an example. (b) Schematic cross-section through the electrode region (AA0) showing interdigitated driven gate (DG) and oating gate (FG) electrodes, and the electric eld coupling them through the dielectric (C60) sample.

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

129

Figure 3-7: Cross section of microdielectrometer sensor. The silicon dioxide insulator is much thinner than the electrode spacing. Y(!) is the comb electrode admittance; CL is the capacitance between the oating electrode and the substrate 3]. a sinusoidal signal is applied to the driven electrode, the corresponding sinusoidally varying voltage that appears on the oating gate can be measured independent of the transistor ampli cation factor. This oating-gate voltage has an amplitude and phase relative to the drive signal that depends on the geometry of the device and on the dielectric properties of the sample medium. The fact that the sensing electrode is electrically oating, combined with the proximity of the ampli er to that electrode, means that a good signal-to-noise ratio can be achieved at very low frequencies (down to 10 3 Hz). As the device geometry is both stable and reproducible, the task of calibrating the measurement reduces to relating the measured amplitude and phase for a particular device geometry to the speci c values of "0 and "00 for the sample. To derive this relationship, consider a cross section of the microdielectrometer sensor shown in Fig. 3-7. The cross section shows the electrodes separated from a conducting ground plane (the silicon substrate) by a silicon dioxide insulating layer whole thickness is much less than the electrode spacing. One of the electrodes is driven with a signal, while the other is connected to the input gate of one of the eld-e ect transistors,

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

130

Figure 3-8: Calibration of the microdielectric sensor, showing contours of constant permittivity and loss factor as a function of the measured magnitude and phase 230, 231]. as described above. Except for a capacitance CL between it and the ground plane, this latter electrode is electrically oating. The capacitance CL integrates the current reaching the oating electrode through the comb electrode admittance Y(!), and develops a voltage which depends on the charge rather than the current 3]. Therefore, instead of providing a direct measurement of Y(!), the microdielectrometer measures this voltage, or, equivalently, the complex transfer function de ned as follows:

Y( H(!) = 1 + j!C!)Y(!) :
L

(3.7)

Like the admittance Y(!), the transfer function H(!) has a magnitude and phase which can be interpreted in terms of an assumed homogeneous dielectric medium with permittivity "0 and loss factor "00, in accordance with a particular model for admittance Y(!) appropriate for the material under study. Figure 3-8 illustrates such a calibration as a contour plot, for the case of Y(!) corresponding to the simple model of Fig. 3-2.

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

131

Figure 3-9: Schematic of \ribbon cable" packaging of the Micromet Instruments' integrated sensor (such as the one shown in Fig. 3-6.) The bonds from the package to the chip are completely passivated, and thus are not exposed either to the material under test or to the ambient environment. Two types of sensors have been used in this study, both supplied by Micromet Instruments Inc. (Newton, MA). The rst, used for characterization of the dielectric properties of C60 lms as function of oxygen di usion (see Sec 4.4.2), was an integrated sensor shown in Fig. 3-6, having an electrode spacing of 12.5 m, and oxide (SiO2) thickness of approximately 1.5 m, and approximately 7.5 cm of total electrode meander (see Table 3.1). The sensors of this type are preassembled in a \ribbon cable" package, as shown in Fig. 3-9. The other type of sensor, the so-called \Micron Sensor" that was introduced by Micromet Instruments in 1995, was used for transport and photoconductivity studies of the C60 lms. It is comprised of two chromium interdigitated electrodes deposited on a quartz substrate (Fig. 3-10). Unlike the sensor in Fig. 3-6, the new Micron Sensor does not have the integrated on-chip electronics, but compensates for the loss of ampli cation with a more closely packed geometry, featuring 1 m interelectrode spacing and a 1 m meander length (see Table 3.1). It is con gured so that a variety of material deposition techniques, including spin-coating, spraying, printing, and thermal sublimation, may be used to apply lms to the surface of the sensor. A measurement system that determines the real and imaginary parts of the di-

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

132

0.75 mm

0.65 mm

3.08 mm

Figure 3-10: The Micron Sensor geometry. Chromium electrodes (1500 A) are deposited on quartz substrate. Courtesy of Huan Lee (Micromet Instruments, Newton, MA) and David Day (Auburn International, Denvers, MA).]

Speci cation Sensor Dimensions Active Face Dimensions Electrode Spacing D Meander Length L Electrode Thickness H Permittivity Range Log Loss Factor Range Log Conductivity Range (S/cm) A/D Ratio Temperature Range

Integrated sensor 1400 0:37500 0:0200 0:200 0:100 12.5 m 7.5 cm N/A 1 to 40 {2 to 3 {16 to {5 N/A {100 to 250 C

Micron sensor 1 1 0:04 cm 3.08 0.75 mm 1 m 1m 1500 A (chromium) 1 to 20 {2 to 2 {13 to {5.5 35 {150 to 200 C

Table 3.1: Microdielectric sensor speci cations. (Source: Micromet Instruments, Inc.)

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

133

Figure 3-11: Functional block diagram of sampled data system used in microdielectrometry measurements 231]. electric constant of the material coating the sensor consists of a frequency response analyzer (similar to the one described in Sec. 3.2.3) and a device (that can be implemented either in hardware on in software) for converting the measured gain and phase into the real and imaginary parts of the dielectric constant. A calibration table (di erent for each type of sensor), such as the one shown in Fig. 3-8, is used for the conversion. For most of our sensor-based dielectric measurements, we used the commercially available Eumetric R System II Microdielectrometer by Micromet Instruments, Inc. (Newton, MA) that combines both the FRA and the lookup functions in a single apparatus. It contains a digital correlator 232] that supplies the sinusoidal drive signal to the driven gate electrode and determines the relative amplitude and phase of the comparator output signal. A functional block diagram of the digital correlator is shown in Fig. 3-11. A master clock synchronizes the digitally synthesized drive

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

134

waveform and the sampling of the comparator output. A discrete Fourier transform (implemented in software) is used to determine the amplitude and phase of the sampled waveform. Based on the accuracy of the amplitude and phase measurement electronics, the "00 sensitivity of the microdielectrometer is about 0.01 232], which for a medium having a dielectric permittivity of 4 corresponds to the loss tangent D sensitivity of less than 0.003. At a frequency of 0.01 Hz, an "00 sensitivity of 0.01 corresponds to a conductivity sensitivity of about 10 16 S/cm. The microdielectrometer was interfaced with an IBM-compatible computer via RS-232 serial port using a versatile LabView software presented in Appendix A.3. Both the integrated sensor and the Micron Sensor include an on-chip temperature indicator. Since all dielectric and transport properties are temperature dependent, the ability to make a temperature measurement at the same point as the dielectric measurement is a useful feature of the microdielectrometry technique. Together with other unique characteristics of this technique described above (e.g., open-faced electrodes, xed calibration, good sensitivity), microdielectrometry is particularly appropriate for performing in-situ measurements, investigating the dynamics of di usion and adsorption, and monitoring the phase transition and cure processes. The results obtained by using this technique for the dielectric and transport studies of C60 lms are presented in Sections 4.7 and 4.4.2, respectively.

3.4 Transport Measurements

As detailed in Sec. 2.3.4, an independent measurement of the DC conductivity 0 is necessary in order to recover the dielectric function "(!) from the e ective dielectric function "e (!) obtained from an AC dielectric measurement. In addition, there are a number of situations where a transport measurement itself provides valuable information about the structure of the sample. For instance, since the resistivity of fullerites has been shown to be extremely sensitive to the presence of oxygen in the

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

135

interstitials of their fcc lattice (see Sec. 4.4.2), transport measurements can be an e ective tool for ascertaining the oxygen content of C60 lms. We have seen in Chapter 1 that, in their pristine state, fullerene solids are insulating. The resistivity of pure crystalline C60 is 107 cm (see Table 1.2), and it can rise by as much as three orders of magnitude as a result of oxygen adsorption. It can easily be shown using Table 3.1 and Fig. 3-10 that, if a Micron sensor is used to measure the resistivity of such an oxygenated C60 lm, the DC resistance being measured will exceed 1 G . Measuring resistances of this magnitude reliably requires special techniques, which are considered below. As always in high-resistance measurements, a known voltage should be applied to the sample, and the resulting current should be measured. This approach is commonly referred to as SVMC (Source Voltage Measure Current).2 The na ve way of measuring an unknown large resistance Rs is to terminate the circuit with a well-calibrated resistor R0 of known value, and to amplify the resulting voltage with a voltage preampli er, as shown in Fig. 3-12(a). In this con guration, R0 serves both as a reference resistor whose value is known precisely, and as a currentlimiting component of the circuit ensuring that, regardless of the value of the test resistor Rs (which may change by many orders of magnitude, as a result of, e.g., oxygen di usion and temperature changes) the current in the circuit is limited to Vs =R0. Using a simple voltage divider formula and solving for Rs , we derive the formula for calculating the unknown resistance Rs from the circuit parameters of Fig. 3-12(a): V Rs = R0 G V s 1 ; (3.8)
m

where G is the di erential ampli er gain, and Vm is the measured voltage. While, under certain conditions, this method may yield satisfactory results, it has a number of drawbacks. Most importantly, to get a large voltage from the current Is, a large
Note that, when dealing with low-resistance (high-conductivity) samples, the Source Current Measure Voltage (SCMV) approach should instead be used.
2

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

136

resistor R0 is necessary,3 which, in combination with cable capacitance and other stray capacitance (see below), can lead to unacceptable penalties in frequency response and phase accuracy. A much better amplitude and phase accuracy in the presence of stray capacitance can be achieved by using a current ampli er, as shown in Fig. 3-12(b). Assuming an ideal (in nite input impedance) linear OpAmp, the unknown resistance Rs is simply

V Rs = Rf V s ; m

(3.9)

where Rf is the feedback resistor. For our measurements, we used the SR570 Low Noise Current Ampli er from Stanford Research (Sunnyvale, CA). A block diagram of the SR570 ampli er is reproduced in Fig. 3-13. A few special consideration for high-resistance measurements 233] are outlined below:

Settling time. The settling time of the circuit is particularly important when making high-resistance measurements. To illustrate, suppose that a 10 G resistive sample is connected to the measuring instrument with a coaxial cable. Considering that a typical coaxial cable has capacitance of 100 pF/m, the circuit's settling time will be 10 G 100 pF/m = 1 s. ments, its input impedance must be considered. If an ohmmeter with input resistance RI is connected to the resistance under test Rs , then the indicated (measured) resistance Rm is

Instrument Loading. Whenever an ohmmeter is used for high resistance measure-

Rm = Rs R RI R : s+ I
In the instances when the con guration depicted in Fig. 3-12(a) was used (in particular, in all transport experiments conducted at MIT), the high precision lm resistors from Caddock Electronics, Inc. (Roseburg, Oregon) were utilized. Available in denominations as high as 100 M , these resistors are remarkably robust both with respect to temperature uctuations and in their frequency response.
3

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

137

Rs Is Vs

HI

R0

V
voltage source

LO

current measurement circuit

(a)

Rf Rs Is Vs
HI

voltage source

LO

current measurement circuit

(b)

Figure 3-12: (a) A schematic of the current measurement circuit using a large terminating resistor R0. The voltage drop over R0 is ampli ed by a di erential OpAmp of gain G, and measured using a voltmeter Vm . (b) A schematic of the current measurement circuit using a current ampli er. The current to voltage conversion gain is controlled by the feedback resistor Rf .

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

138

Figure 3-13: A block diagram of the SR570 Low Noise Current Ampli er. The input current can be o set to suppress any undesired background currents. The voltage at the \+" terminal of the ampli er is controlled by an adjustable 5 V bias source (0 V at the \+" terminal makes the SR570's input a virtual null). The current to voltage conversion gain is controlled by the feedback resistor Rf . Sensitivities from 1 pA/V to 1 mA/V can be selected in a 1-2-5 sequence. As the sensitivity increases and Rf gets larger, any capacitance present will cause the bandwidth of the instrument to su er. The SR570 provides a choice between a \high bandwidth" mode, where Rf is chosen as small as possible and the gain is made up later in the ampli er chain, and a \low noise" mode, where Rf is chosen as large as possible to optimize noise performance. The noise performance and frequency response of the instrument can be further adjusted using two identical rst order RC lters that can be con gured as lowpass, highpass, or bandpass, with the cuto frequencies controlled from the front panel of the instrument.

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

139

Thus, if an ohmmeter with input resistance of 1 G is used to measure a 100 M sample, the resulting measurement will be o by 9%.

Guarding. Guarding high-resistance test connections can signi cantly reduce the

e ects of leakage resistance, increase measurement accuracy, and improve time response (bandwidth) of the instrument. Under this technique, the measurement cable shield is driven with a unity gain ampli er to maintain the same potential as the wire inside the shield. Since both the shield and the wire are then at the same potential, the leakage current is eliminated. Most modern instruments provide a guard terminal facilitating this technique.

Leakage Currents. In the SVMC con gurations, such as the one pictured in Fig. 3-

12(b), care should be taken to suppress any leakage currents present in the system. The baseline suppression feature of the SR570 ampli er (see Fig. 313) is available to o set the input current so as to suppress any undesired background currents.

Provided that the di culties outlined above are addressed and a successful measurement has been performed, the output of the current ampli er (ranging from -5 V to +5 V for SR570) needs to be converted to the resistivity of the sample being measured. For samples in the parallel-plate con guration (Fig. 3-1), this is merely a matter of plugging in the electrode area A and separation distance d into the formula = Rs A ; d (3.10)

where Rs is the sample resistance calculated using Eqn. (3.9). In practice, instead of the Eqn. (3.9), ampli er gain Gamp and output voltage Vamp are used to calculate Rs:

Rs = G Vs : ampVamp
Note that the ampli er gain Gamp is expressed in amperes per volt (A/V).

(3.11)

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

140

The situation gets somewhat more complex when the Micron sensor (see Sec. 3.3) is utilized to measure the resistivity. The sensor's geometrical parameters listed in Table 3.1 should then be used to calculate the geometrical factor A=d of Eqn. (3.10). For lms thicker than the sensor electrode thickness H ,

A=L H ; d D

(3.12)

where L and H are, respectively, the meander length and the electrode spacing of the sensor. Given the parameters in Table 3.1, we calculate

A = 15cm: d
Clearly for lms thinner than H , the lm thickness d lm should be used in place of H in Eqn. (3.12). Note that, when the Micron sensor is used to measure the dielectric properties of materials, the geometrical parameter A=d can be used to calculate the dielectric constant "0 from the measured capacitance Cm:

Cm "0 = A=d :

(3.13)

However, the simple-minded way of calculating A=d utilized in Eqn. (3.12) is no longer adequate, since it does not take into account the fringing elds imposed by the geometry of the sensor (see Figs. 3-6 and 3-10), which cannot be neglected in AC measurements. Micromet Instruments recommends using A=d = 35 cm, a value determined from experimental calibrations. In assessing the merits of this value relative to the approximation of Eqn (3.12), it is important to keep in mind that, in dielectric measurements, the fringe elds extend not only above the surface of the sensor into the bulk of the measured material, but also into the quartz substrate, whose dielectric constant "0quartz = 3:75 contributes additional polarization (which we

CHAPTER 3. LOW FREQUENCY DIELECTRIC MEASUREMENTS

141

consider to be frequency-independent) that gets internalized in the A=d geometrical parameter.